EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Monocular Depth Estimation from a Fisheye Camera Based on Knowledge Distillation

Eunjin Son iD, Jiho Choi iD, Jimin Song iD, Yongsik Jin iD, Sang Jun Lee iD

Sensors · 16.12.2023 · 23 (24) · 9866

DOI: 10.3390/s23249866 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sensors, online erschienen am 16.12.2023. Verfasst von Eunjin Son und 4 weiteren Autor:innen. Beteiligte Einrichtung: Division of Electronic Engineering, Jeonbuk National University, 567 Baekje-daero, Deokjin-gu, Jeonju 54896, Republic of Korea. Die Arbeit wurde bisher 12-mal zitiert (Crossref) und nennt 60 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Ministry of Trade, Industry and Energy, Electronics and Telecommunications Research Institute, the Ministry of Trade, Industry & Energy (MOTIE, Korea).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Monocular depth estimation is a task aimed at predicting pixel-level distances from a single RGB image. This task holds significance in various applications including autonomous driving and robotics. In particular, the recognition of surrounding environments is important to avoid collisions during autonomous parking. Fisheye cameras are adequate to acquire visual information from a wide field of view, reducing blind spots and preventing potential collisions. While there have been increasing demands for fisheye cameras in visual-recognition systems, existing research on depth estimation has primarily focused on pinhole camera images. Moreover, depth estimation from fisheye images poses additional challenges due to strong distortion and the lack of public datasets. In this work, we propose a novel underground parking lot dataset called JBNU-Depth360, which consists of fisheye camera images and their corresponding LiDAR projections. Our proposed dataset was composed of 4221 pairs of fisheye images and their corresponding LiDAR point clouds, which were obtained from six driving sequences. Furthermore, we employed a knowledge-distillation technique to improve the performance of the state-of-the-art depth-estimation models. The teacher–student learning framework allows the neural network to leverage the information in dense depth predictions and sparse LiDAR projections. Experiments were conducted on the KITTI-360 and JBNU-Depth360 datasets for analyzing the performance of existing depth-estimation models on fisheye camera images. By utilizing the self-distillation technique, the AbsRel and SILog error metrics were reduced by 1.81% and 1.55% on the JBNU-Depth360 dataset. The experimental results demonstrated that the self-distillation technique is beneficial to improve the performance of depth-estimation models.

Schlagwörter

Stichwörter der Autor:innen

fisheye cameraknowledge distillationmonocular depth estimationparking lot datasetsupervised depth estimation

Autor:innen und Institutionen

  • Division of Electronic Engineering, Jeonbuk National University, 567 Baekje-daero, Deokjin-gu, Jeonju 54896, Republic of Korea.
  • IT Convergence Research Section, Electronics and Telecommunications Research Institute (ETRI), Daegu 42995, Republic of Korea.

ORCID-Profile: Eunjin Son, Jiho Choi, Jimin Song, Yongsik Jin, Sang Jun Lee

Förderung

  • Ministry of Trade, Industry and Energy
  • Electronics and Telecommunications Research Institute
  • the Ministry of Trade, Industry & Energy (MOTIE, Korea)
  • the Electronics and Telecommunications Research Institute (ETRI)

Zitieren

APA

Son, E., Choi, J., Song, J., Jin, Y., & Lee, S. J. (2023). Monocular Depth Estimation from a Fisheye Camera Based on Knowledge Distillation. Sensors, 23(24), 9866. https://doi.org/10.3390/s23249866

Vancouver

Son E, Choi J, Song J, Jin Y, Lee SJ. Monocular Depth Estimation from a Fisheye Camera Based on Knowledge Distillation. Sensors. 2023;23(24):9866. doi:10.3390/s23249866

BibTeX
@article{Son2023,
  author = {Son, Eunjin and Choi, Jiho and Song, Jimin and Jin, Yongsik and Lee, Sang Jun},
  title = {Monocular Depth Estimation from a Fisheye Camera Based on Knowledge Distillation},
  journal = {Sensors},
  year = {2023},
  volume = {23},
  number = {24},
  pages = {9866},
  doi = {10.3390/s23249866}
}
Kostenlos starten →